[ntp:questions] Clock jumps when refclock used

A C agcarver+ntp at acarver.net
Sat Apr 14 07:43:37 UTC 2012


On 4/13/2012 23:35, A C wrote:
> On 4/13/2012 23:26, Dave Hart wrote:
>> Please share the sys_fuzz messages you see along with any insight into
>> how they differ with configuration that you can provide. On-list or
>> off-list is fine. Those messages are indicating time apparently going
>> backwards and may be revealing a bug in ntpd or your OS.
>>
>> Cheers,
>> Dave Hart
>>
>
> I will do so. I get quite a few of them in the refclock configuration. I
> get some with Internet only but not as many. I'm going to let the
> current configuration run for a few more days and then provide those
> sys_fuzz messages. Afterwards I'll switch the configuration and track it
> again. The current configuration has both refclocks noselected, kernel
> discipline disabled, and five Internet sources. There are no preferred
> peers or any other unusual configurations other than a minpoll 6 on the
> Internet servers (left over configuration from the refclocks). The
> polling period is moving so the clock selection is working. Right now
> the polling period has set itself for 1024 seconds with an average
> offset of about 4ms.

I did notice this in the current Internet only configuration, one 
Internet clock went crazy just now:

peerstats:
56031 27012.604 130.207.165.28  9374  99542.105309349  0.000122070 
0.015549461  99542.097097054



ntpq line:
-130.207.165.28  130.207.244.240  2 u  190 1024  377    0.122  9954210 
9954209

There was a fuzz shortly afterwards (the above was at 07:30:12) :
14 Apr 07:30:48 : ts_min 1334388647.362400444 ts_prev 
1334388647.362295452 ts 1334388648.362303446
14 Apr 07:30:48 : sys_fuzz 104992 nsec, this fuzz 0.000053869, prior 
0.000074835
14 Apr 07:30:48 : get_systime prev result 0xd333a628.5d03feac is 
0.000984891 later than 0xd333a628.5cc372f1


Note the delay, it seems rather curious that the delay is exactly the 
computed jitter of my system.  That's the same value used in the jitter 
field of my PPS source:

peerstats:
56031 27118.381 127.127.22.0    9024  0.004945328  0.000000000 
0.000353156  0.000122070
ntpq:
127.127.22.0    .PPS.            0 l    9   16  377    0.000    5.019 0.122


Coincidence or did something accidentally jumble some memory addresses?


More information about the questions mailing list